Call:
glm(formula = MondayOpenComparedToFridayCloseChangeInPercent ~
FridayCloseComparedToThursdayCloseChangeInPercent + FridayIntradayChangeInPercent,
family = gaussian, data = df)
Coefficients:
Estimate Std. Error t value
(Intercept) 0.01566 0.02757 0.568
FridayCloseComparedToThursdayCloseChangeInPercent -0.11464 0.03659 -3.133
FridayIntradayChangeInPercent 0.12679 0.04563 2.779
Pr(>|t|)
(Intercept) 0.57003
FridayCloseComparedToThursdayCloseChangeInPercent 0.00178 **
FridayIntradayChangeInPercent 0.00557 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for gaussian family taken to be 0.6980752)
Null deviance: 646.62 on 918 degrees of freedom
Residual deviance: 639.44 on 916 degrees of freedom
AIC: 2282.7
Number of Fisher Scoring iterations: 2analysis
Due 12/18/2023
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose,
family = binomial, data = df2)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.34277 0.09719 3.527 0.000421 ***
didFridayCloseHigherThanThursdayClose -0.32206 0.13315 -2.419 0.015572 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1261.3 on 917 degrees of freedom
AIC: 1265.3
Number of Fisher Scoring iterations: 4
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ wasFridayIntradayPositive,
family = binomial, data = df2)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.33647 0.09626 3.495 0.000473 ***
wasFridayIntradayPositive -0.31542 0.13300 -2.372 0.017711 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1261.6 on 917 degrees of freedom
AIC: 1265.6
Number of Fisher Scoring iterations: 4
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
wasFridayIntradayPositive * fridayWasATripleWitchDay, family = binomial,
data = df3)
Coefficients:
Estimate Std. Error z value
(Intercept) 0.4115 0.1110 3.707
didFridayCloseHigherThanThursdayClose -0.2196 0.1748 -1.256
wasFridayIntradayPositive -0.2080 0.1771 -1.175
fridayWasATripleWitchDay -0.2341 0.2832 -0.827
wasFridayIntradayPositive:fridayWasATripleWitchDay 0.2518 0.5287 0.476
Pr(>|z|)
(Intercept) 0.00021 ***
didFridayCloseHigherThanThursdayClose 0.20904
wasFridayIntradayPositive 0.24008
fridayWasATripleWitchDay 0.40840
wasFridayIntradayPositive:fridayWasATripleWitchDay 0.63397
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1259.5 on 914 degrees of freedom
AIC: 1269.5
Number of Fisher Scoring iterations: 4# A tibble: 2 × 4
SpecialCondition Ones Zeros Ratio
<chr> <int> <int> <dbl>
1 Condition Met 35 31 1.13
2 Other 464 389 1.19MONDAY OPEN TO FRIDAY CLOSE RATIO DENSITY PLOT & MONDAY CLOSE TO FRIDAY CLOSE RATIO DENSITY PLOT
df4 <- read.csv('data/addingFOMC/SPY_ready_for_analysis_triple_witch_NFP_CSI_FOMC_with_month.csv', header = TRUE)
df4$MondayCloseToFridayCloseRatio <- df4$FollowingWeekMondayClose / df4$PrecedingWeekFridayClose
ggplot(df4, aes(x = MondayOpenToFridayCloseRatio)) +
geom_density(fill = "blue", alpha = 0.5) +
scale_x_continuous(trans = 'log', limits = c(0.95, 1.05)) +
labs(title = "Density of Monday Open to Friday Close Ratio",
x = "Monday Open to Friday Close Ratio",
y = "Density")
ggplot(df4, aes(x = MondayCloseToFridayCloseRatio)) +
geom_density(fill = "pink", alpha = 0.5) +
scale_x_continuous(trans = 'log', limits = c(0.95, 1.05)) +
labs(title = "Density of Monday Close to Friday Close Ratio",
x = "Monday Close to Friday Close Ratio",
y = "Density")
WasMondayOpenHigherThanFridayClose and WasMondayCloseHigherThanFridayClose BREAKDOWN
#1 vs 0 proportion break down table with percentages for df4$WasMondayOpenHigherThanFridayClose and df4$WasMondayCloseHigherThanFridayClose
# Count occurrences for WasMondayOpenHigherThanFridayClose
table_open <- table(df4$WasMondayOpenHigherThanFridayClose)
# Convert counts to proportions and then to percentages for WasMondayOpenHigherThanFridayClose
percentage_open <- prop.table(table_open) * 100
# Count occurrences for WasMondayCloseHigherThanFridayClose
table_close <- table(df4$WasMondayCloseHigherThanFridayClose)
# Convert counts to proportions and then to percentages for WasMondayCloseHigherThanFridayClose
percentage_close <- prop.table(table_close) * 100
formatted_yes_open <- sprintf("%.1f%%", percentage_open[2])
formatted_no_open <- sprintf("%.1f%%", percentage_open[1])
formatted_yes_close <- sprintf("%.1f%%", percentage_close[2])
formatted_no_close <- sprintf("%.1f%%", percentage_close[1])
breakdown_table <- data.frame(
Category = c('Monday Open Price > Friday Close Price', 'Monday Close Price > Friday Close Price'),
Yes = c(formatted_yes_open, formatted_yes_close),
No = c(formatted_no_open, formatted_no_close)
)
library(knitr)
# Print the table
kable(breakdown_table, caption = "Proportions of Monday Open and Close vs. Friday Close", format = "html")| Category | Yes | No |
|---|---|---|
| Monday Open Price > Friday Close Price | 54.3% | 45.7% |
| Monday Close Price > Friday Close Price | 54.2% | 45.8% |
WasMondayOpenHigherThanFridayClose vs NFPOutcomeIsShock BREAKDOWN
df4$NFPOutcomeIsShock <- ifelse(df4$NFPOutcome == "Shock", 1, 0)
#1. breakdown table of WasMondayOpenHigherThanFridayClose (1 vs 0) and NFPOutcomeIsShock (1 vs 0)
breakdown_table_monday_open_vs_nfp_shock <- table(df4$WasMondayOpenHigherThanFridayClose, df4$NFPOutcomeIsShock)
rownames(breakdown_table_monday_open_vs_nfp_shock) <- c("Monday Open Lower Than Friday Close", "Monday Open Higher Than Friday Close")
colnames(breakdown_table_monday_open_vs_nfp_shock) <- c("Nonfarm Payrolls Non-shock", "Nonfarm Payrolls Shock")
breakdown_table_column_percentages <- prop.table(breakdown_table_monday_open_vs_nfp_shock, 2) * 100
#print(breakdown_table_column_percentages)
breakdown_table_column_percentages_df <- as.data.frame.matrix(breakdown_table_column_percentages)
# Format each value to one decimal place with a percentage sign
breakdown_table_column_percentages_df_formatted <- data.frame(lapply(breakdown_table_column_percentages_df, function(x) sprintf("%.1f%%", x)))
# Print the formatted dataframe
#print(breakdown_table_column_percentages_df_formatted)
breakdown_table_WasMondayOpenHigherThanFridayClose_vs_NFPOutcomeIsShock <- data.frame(
. = c("Monday Open Lower Than Friday Close", "Monday Open Higher Than Friday Close"),
NFP_Nonshock = c(breakdown_table_column_percentages_df_formatted[1][[1]][1], breakdown_table_column_percentages_df_formatted[1][[1]][2]),
NFP_Shock = c(breakdown_table_column_percentages_df_formatted[2][[1]][1], breakdown_table_column_percentages_df_formatted[2][[1]][2])
)
library(knitr)
# Print the table
kable(breakdown_table_WasMondayOpenHigherThanFridayClose_vs_NFPOutcomeIsShock, caption = "Proportion Breakdown of Monday Open Higher Than Friday Close vs Nonfarm Payroll Shock", format = "html")| . | NFP_Nonshock | NFP_Shock |
|---|---|---|
| Monday Open Lower Than Friday Close | 45.8% | 41.7% |
| Monday Open Higher Than Friday Close | 54.2% | 58.3% |
WasMondayCloseHigherThanFridayClose vs NFPOutcomeIsShock BREAKDOWN
df4$NFPOutcomeIsShock <- ifelse(df4$NFPOutcome == "Shock", 1, 0)
#1. breakdown table of WasMondayCloseHigherThanFridayClose (1 vs 0) and NFPOutcomeIsShock (1 vs 0)
breakdown_table_monday_close_vs_nfp_shock <- table(df4$WasMondayCloseHigherThanFridayClose, df4$NFPOutcomeIsShock)
rownames(breakdown_table_monday_close_vs_nfp_shock) <- c("Monday Close Lower Than Friday Close", "Monday Close Higher Than Friday Close")
colnames(breakdown_table_monday_close_vs_nfp_shock) <- c("Nonfarm Payrolls Non-shock", "Nonfarm Payrolls Shock")
breakdown_table_column_percentages2 <- prop.table(breakdown_table_monday_close_vs_nfp_shock, 2) * 100
#print(breakdown_table_column_percentages)
breakdown_table_column_percentages_df2 <- as.data.frame.matrix(breakdown_table_column_percentages2)
# Format each value to one decimal place with a percentage sign
breakdown_table_column_percentages_df_formatted2 <- data.frame(lapply(breakdown_table_column_percentages_df2, function(x) sprintf("%.1f%%", x)))
# Print the formatted dataframe
#print(breakdown_table_column_percentages_df_formatted)
breakdown_table_WasMondayCloseHigherThanFridayClose_vs_NFPOutcomeIsShock <- data.frame(
. = c("Monday Open Lower Than Friday Close", "Monday Open Higher Than Friday Close"),
NFP_Nonshock = c(breakdown_table_column_percentages_df_formatted2[1][[1]][1], breakdown_table_column_percentages_df_formatted2[1][[1]][2]),
NFP_Shock = c(breakdown_table_column_percentages_df_formatted2[2][[1]][1], breakdown_table_column_percentages_df_formatted2[2][[1]][2])
)
library(knitr)
# Print the table
kable(breakdown_table_WasMondayCloseHigherThanFridayClose_vs_NFPOutcomeIsShock, caption = "Proportion Breakdown of Monday Close Higher Than Friday Close vs Nonfarm Payroll Shock", format = "html")| . | NFP_Nonshock | NFP_Shock |
|---|---|---|
| Monday Open Lower Than Friday Close | 45.4% | 75.0% |
| Monday Open Higher Than Friday Close | 54.6% | 25.0% |
Warning: `aes_string()` was deprecated in ggplot2 3.0.0.
ℹ Please use tidy evaluation idioms with `aes()`.
ℹ See also `vignette("ggplot2-in-packages")` for more information.
Warning: Removed 614 rows containing missing values (`geom_point()`).
[[1]]
[[2]]
[[3]]
[[4]]
[[5]]
[[6]]
[[7]]
[[8]]
[[9]]
[[10]]
[[11]]
[[12]]
[[13]]
[[14]]
[[15]]
[[16]]
[[17]]
[[18]]
[[19]]
[[20]]
[[21]]
[[22]]
[[23]]
[[24]]
[[25]]
[[26]]
[[27]]
[[28]]
[[29]]
[[30]]
[[31]]
[[32]]
[[33]]
[[34]]
[[35]]
[[36]]
[[37]]Warning: Removed 614 rows containing missing values (`geom_point()`).
[[38]]
[[39]]
[[40]]
[[41]]
[[42]]
[[43]]
[[44]]
[[45]]
[[46]]
df6 <- read.csv('data/addingUmichSentiment/SPY_ready_for_analysis_triple_witch_NFP_CSI.csv', header = TRUE)
summary(df6$CSIActualMinusExpected) Min. 1st Qu. Median Mean 3rd Qu. Max. NA's
-11.0000 -1.4000 0.1000 -0.2161 1.2000 9.1000 614 table(df6$WasMondayOpenHigherThanFridayClose)
0 1
420 499 boxplot(CSIActualMinusExpected ~ WasMondayOpenHigherThanFridayClose, data = df6, main = "CSIActualMinusExpected vs WasMondayOpenHigherThanFridayClose")
library(ggplot2)
ggplot(df6, aes(x = CSIActualMinusExpected, fill = factor(WasMondayOpenHigherThanFridayClose))) +
geom_density(alpha = 0.5) +
labs(fill = "WasMondayOpenHigher")Warning: Removed 614 rows containing non-finite values (`stat_density()`).
# T-test or Mann-Whitney U Test (depending on data distribution)
t.test(CSIActualMinusExpected ~ WasMondayOpenHigherThanFridayClose, data = df6)
Welch Two Sample t-test
data: CSIActualMinusExpected by WasMondayOpenHigherThanFridayClose
t = -0.56092, df = 287.94, p-value = 0.5753
alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
95 percent confidence interval:
-0.8476626 0.4716685
sample estimates:
mean in group 0 mean in group 1
-0.3128378 -0.1248408 df6$CSIOutcomeIsShock <- ifelse(df6$CSIOutcome == "Shock", 1, 0)
table(df6$CSIOutcomeIsShock, df6$WasMondayOpenHigherThanFridayClose)
0 1
0 404 489
1 16 10chisq.test(table(df6$CSIOutcomeIsShock, df6$WasMondayOpenHigherThanFridayClose))
Pearson's Chi-squared test with Yates' continuity correction
data: table(df6$CSIOutcomeIsShock, df6$WasMondayOpenHigherThanFridayClose)
X-squared = 2.0873, df = 1, p-value = 0.1485df6 <- read.csv('data/addingUmichSentiment/SPY_ready_for_analysis_triple_witch_NFP_CSI.csv', header = TRUE)
df6$NFPOutcomeIsShock <- ifelse(df6$NFPOutcome == "Shock", 1, 0)
df6$CSIOutcomeIsShock <- ifelse(df6$CSIOutcome == "Shock", 1, 0)
# Binomial model
model6 <- glm(WasMondayCloseHigherThanFridayClose ~ wasFridayIntradayPositive + NFPOutcomeIsShock + fridayWasATripleWitchDay + CSIOutcomeIsShock,
data = df6, family = binomial)
# Print the summary of the model
summary(model6)
Call:
glm(formula = WasMondayCloseHigherThanFridayClose ~ wasFridayIntradayPositive +
NFPOutcomeIsShock + fridayWasATripleWitchDay + CSIOutcomeIsShock,
family = binomial, data = df6)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.32631 0.10244 3.185 0.00145 **
wasFridayIntradayPositive -0.18199 0.13519 -1.346 0.17825
NFPOutcomeIsShock -1.28997 0.67133 -1.922 0.05467 .
fridayWasATripleWitchDay -0.52945 0.23946 -2.211 0.02703 *
CSIOutcomeIsShock -0.06743 0.40024 -0.168 0.86621
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.5 on 918 degrees of freedom
Residual deviance: 1257.3 on 914 degrees of freedom
AIC: 1267.3
Number of Fisher Scoring iterations: 4df7 <- read.csv('data/addingFOMC/SPY_ready_for_analysis_triple_witch_NFP_CSI_FOMC.csv', header = TRUE)
df7$NFPOutcomeIsShock <- ifelse(df7$NFPOutcome == "Shock", 1, 0)
df7$CSIOutcomeIsShock <- ifelse(df7$CSIOutcome == "Shock", 1, 0)
df7_filtered <- rbind(df7[1:220, ], df7[264:nrow(df7), ])
df7_filtered$NFPOutcomeIsShock <- ifelse(df7_filtered$NFPOutcome == "Shock", 1, 0)
df7_filtered$CSIOutcomeIsShock <- ifelse(df7_filtered$CSIOutcome == "Shock", 1, 0)
df7_filtered$PrecedingWeekFridayVolumeMINUSPrecedingWeekThursdayVolume <- (df7_filtered$PrecedingWeekFridayVolume) - (df7_filtered$PrecedingWeekThursdayVolume)
#df7$FridayCloseToFridayOpenRatio <- (df$PrecedingWeekFridayClose) / (df$PrecedingWeekFridayOpen)
model7BUYFRICLOSESELLMONOPEN <- glm(WasMondayOpenHigherThanFridayClose ~ wasFridayIntradayPositive,
data = df7_filtered, family = binomial)
summary(model7BUYFRICLOSESELLMONOPEN)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ wasFridayIntradayPositive,
family = binomial, data = df7_filtered)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.33647 0.09897 3.400 0.000675 ***
wasFridayIntradayPositive -0.32770 0.13626 -2.405 0.016178 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1208.5 on 875 degrees of freedom
Residual deviance: 1202.7 on 874 degrees of freedom
AIC: 1206.7
Number of Fisher Scoring iterations: 4# wasFridayIntradayPositive
#model7BUYFRICLOSESELLMONclose <- glm(WasMondayCloseHigherThanFridayClose ~ NFPOutcomeIsShock + fridayWasATripleWitchDay,
# data = df7_filtered, family = binomial)
#summary(model7BUYFRICLOSESELLMONclose)df8 <- read.csv('data/addingFOMC/SPY_ready_for_analysis_triple_witch_NFP_CSI_FOMC_with_month.csv', header = TRUE)
#df8 <- rbind(df8[1:220, ], df8[264:nrow(df8), ])
df8$NFPOutcomeIsShock <- ifelse(df8$NFPOutcome == "Shock", 1, 0)
df8$CSIOutcomeIsShock <- ifelse(df8$CSIOutcome == "Shock", 1, 0)
df8$PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO <- (df8$PrecedingWeekFridayVolume) / (df8$PrecedingWeekThursdayVolume)
null_model <- glm(WasMondayOpenHigherThanFridayClose ~ 1,
data = df8, family = binomial)
summary(null_model)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ 1, family = binomial,
data = df8)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.17235 0.06622 2.603 0.00925 **
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1267.2 on 918 degrees of freedom
AIC: 1269.2
Number of Fisher Scoring iterations: 3full_model <- glm(WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose + FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive + FridayIntradayChangeInPercent + fridayWasATripleWitchDay + NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek + FOMCAction + FOMCActualVSExpected + FOMCActionSpecial + NFPOutcomeIsShock + CSIOutcomeIsShock + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO,
data = df8, family = binomial)
summary(full_model)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek +
FOMCAction + FOMCActualVSExpected + FOMCActionSpecial + NFPOutcomeIsShock +
CSIOutcomeIsShock + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO,
family = binomial, data = df8)
Coefficients: (7 not defined because of singularities)
Estimate Std. Error
(Intercept) 0.79139 0.27674
didFridayCloseHigherThanThursdayClose -0.36570 0.21345
FridayCloseComparedToThursdayCloseChangeInPercent 0.05392 0.10829
wasFridayIntradayPositive -0.41402 0.21009
FridayIntradayChangeInPercent 0.19213 0.13439
fridayWasATripleWitchDay -0.28388 0.25759
NFPDayOrNot -0.27942 0.18926
NFPOutcomeShock 0.52127 0.64363
NFPOutcomeUpbeat 0.40145 0.64233
CSIOutcomeBaseline -0.15524 0.15692
CSIOutcomeShock -1.03298 0.39505
CSIOutcomeUpbeat 0.13896 0.65833
CSIDayOrNot NA NA
wasItFOMCWeek -0.83568 0.87603
FOMCActionCut -0.65468 1.42167
FOMCActionHike 1.37253 1.01127
FOMCActionPause 1.11306 0.89877
FOMCActionSuperCut 0.50344 1.68962
FOMCActionSuperHike NA NA
FOMCActualVSExpectedDovish -14.03440 535.41356
FOMCActualVSExpectedHawkish 12.68100 535.41148
FOMCActualVSExpectedMatch NA NA
FOMCActionSpecialSuperCut NA NA
FOMCActionSpecialSuperHike NA NA
NFPOutcomeIsShock NA NA
CSIOutcomeIsShock NA NA
PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO -0.07900 0.20513
z value Pr(>|z|)
(Intercept) 2.860 0.00424 **
didFridayCloseHigherThanThursdayClose -1.713 0.08666 .
FridayCloseComparedToThursdayCloseChangeInPercent 0.498 0.61850
wasFridayIntradayPositive -1.971 0.04876 *
FridayIntradayChangeInPercent 1.430 0.15283
fridayWasATripleWitchDay -1.102 0.27043
NFPDayOrNot -1.476 0.13985
NFPOutcomeShock 0.810 0.41800
NFPOutcomeUpbeat 0.625 0.53198
CSIOutcomeBaseline -0.989 0.32251
CSIOutcomeShock -2.615 0.00893 **
CSIOutcomeUpbeat 0.211 0.83282
CSIDayOrNot NA NA
wasItFOMCWeek -0.954 0.34011
FOMCActionCut -0.460 0.64516
FOMCActionHike 1.357 0.17471
FOMCActionPause 1.238 0.21556
FOMCActionSuperCut 0.298 0.76573
FOMCActionSuperHike NA NA
FOMCActualVSExpectedDovish -0.026 0.97909
FOMCActualVSExpectedHawkish 0.024 0.98110
FOMCActualVSExpectedMatch NA NA
FOMCActionSpecialSuperCut NA NA
FOMCActionSpecialSuperHike NA NA
NFPOutcomeIsShock NA NA
CSIOutcomeIsShock NA NA
PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO -0.385 0.70013
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1236.1 on 899 degrees of freedom
AIC: 1276.1
Number of Fisher Scoring iterations: 12forward_model <- step(null_model,
scope = list(lower = null_model, upper = full_model),
direction = "forward")Start: AIC=1269.21
WasMondayOpenHigherThanFridayClose ~ 1
Df Deviance AIC
+ CSIOutcomeIsShock 1 1260.0 1264.0
+ didFridayCloseHigherThanThursdayClose 1 1261.3 1265.3
+ wasFridayIntradayPositive 1 1261.6 1265.6
+ CSIOutcome 3 1259.6 1267.6
<none> 1267.2 1269.2
+ CSIDayOrNot 1 1265.7 1269.7
+ NFPDayOrNot 1 1265.9 1269.9
+ FridayCloseComparedToThursdayCloseChangeInPercent 1 1266.7 1270.7
+ wasItFOMCWeek 1 1266.9 1270.9
+ FridayIntradayChangeInPercent 1 1267.0 1271.0
+ NFPOutcomeIsShock 1 1267.1 1271.1
+ fridayWasATripleWitchDay 1 1267.2 1271.2
+ PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1267.2 1271.2
+ FOMCActionSpecial 2 1265.6 1271.6
+ FOMCActualVSExpected 3 1264.1 1272.1
+ NFPOutcome 2 1266.7 1272.7
+ FOMCAction 5 1261.1 1273.1
Step: AIC=1264.04
WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock
Df Deviance AIC
+ wasFridayIntradayPositive 1 1254.4 1260.4
+ didFridayCloseHigherThanThursdayClose 1 1254.5 1260.5
<none> 1260.0 1264.0
+ NFPDayOrNot 1 1258.3 1264.3
+ FridayCloseComparedToThursdayCloseChangeInPercent 1 1259.5 1265.5
+ CSIDayOrNot 1 1259.8 1265.8
+ FridayIntradayChangeInPercent 1 1259.8 1265.8
+ wasItFOMCWeek 1 1259.8 1265.8
+ NFPOutcomeIsShock 1 1260.0 1266.0
+ fridayWasATripleWitchDay 1 1260.0 1266.0
+ PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1260.0 1266.0
+ FOMCActionSpecial 2 1258.3 1266.3
+ FOMCActualVSExpected 3 1257.0 1267.0
+ CSIOutcome 2 1259.6 1267.6
+ NFPOutcome 2 1259.7 1267.7
+ FOMCAction 5 1254.0 1268.0
Step: AIC=1260.42
WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock + wasFridayIntradayPositive
Df Deviance AIC
+ FridayIntradayChangeInPercent 1 1251.5 1259.5
<none> 1254.4 1260.4
+ NFPDayOrNot 1 1253.1 1261.1
+ didFridayCloseHigherThanThursdayClose 1 1253.2 1261.2
+ FridayCloseComparedToThursdayCloseChangeInPercent 1 1253.9 1261.9
+ PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1254.1 1262.1
+ fridayWasATripleWitchDay 1 1254.1 1262.1
+ CSIDayOrNot 1 1254.2 1262.2
+ wasItFOMCWeek 1 1254.3 1262.3
+ NFPOutcomeIsShock 1 1254.3 1262.3
+ FOMCActionSpecial 2 1252.8 1262.8
+ FOMCActualVSExpected 3 1251.8 1263.8
+ NFPOutcome 2 1253.9 1263.9
+ CSIOutcome 2 1254.0 1264.0
+ FOMCAction 5 1248.3 1264.3
Step: AIC=1259.54
WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock + wasFridayIntradayPositive +
FridayIntradayChangeInPercent
Df Deviance AIC
+ didFridayCloseHigherThanThursdayClose 1 1249.2 1259.2
<none> 1251.5 1259.5
+ NFPDayOrNot 1 1250.2 1260.2
+ fridayWasATripleWitchDay 1 1251.3 1261.3
+ CSIDayOrNot 1 1251.3 1261.3
+ wasItFOMCWeek 1 1251.4 1261.4
+ FridayCloseComparedToThursdayCloseChangeInPercent 1 1251.4 1261.4
+ PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1251.4 1261.4
+ NFPOutcomeIsShock 1 1251.4 1261.4
+ FOMCActionSpecial 2 1249.6 1261.6
+ FOMCAction 5 1244.6 1262.6
+ FOMCActualVSExpected 3 1248.8 1262.8
+ NFPOutcome 2 1251.0 1263.0
+ CSIOutcome 2 1251.2 1263.2
Step: AIC=1259.19
WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + didFridayCloseHigherThanThursdayClose
Df Deviance AIC
<none> 1249.2 1259.2
+ NFPDayOrNot 1 1247.9 1259.9
+ fridayWasATripleWitchDay 1 1248.7 1260.7
+ PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1248.9 1260.9
+ CSIDayOrNot 1 1248.9 1260.9
+ FridayCloseComparedToThursdayCloseChangeInPercent 1 1248.9 1260.9
+ FOMCActionSpecial 2 1247.1 1261.1
+ wasItFOMCWeek 1 1249.1 1261.1
+ NFPOutcomeIsShock 1 1249.1 1261.1
+ FOMCActualVSExpected 3 1246.4 1262.4
+ FOMCAction 5 1242.5 1262.5
+ NFPOutcome 2 1248.6 1262.6
+ CSIOutcome 2 1248.7 1262.7summary(forward_model)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
didFridayCloseHigherThanThursdayClose, family = binomial,
data = df8)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.5668 0.1313 4.317 1.59e-05 ***
CSIOutcomeIsShock -0.9661 0.3891 -2.483 0.0130 *
wasFridayIntradayPositive -0.4103 0.2047 -2.004 0.0451 *
FridayIntradayChangeInPercent 0.2150 0.1084 1.983 0.0474 *
didFridayCloseHigherThanThursdayClose -0.2738 0.1786 -1.533 0.1252
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1249.2 on 914 degrees of freedom
AIC: 1259.2
Number of Fisher Scoring iterations: 4backward_model <- step(full_model, direction = "backward")Start: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek +
FOMCAction + FOMCActualVSExpected + FOMCActionSpecial + NFPOutcomeIsShock +
CSIOutcomeIsShock + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Step: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek +
FOMCAction + FOMCActualVSExpected + FOMCActionSpecial + NFPOutcomeIsShock +
PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Step: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek +
FOMCAction + FOMCActualVSExpected + FOMCActionSpecial + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Step: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + wasItFOMCWeek +
FOMCAction + FOMCActualVSExpected + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Step: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + CSIDayOrNot + FOMCAction +
FOMCActualVSExpected + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Step: AIC=1276.08
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + NFPOutcome + CSIOutcome + FOMCAction + FOMCActualVSExpected +
PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Df Deviance AIC
- NFPOutcome 2 1237.1 1273.1
- FOMCAction 4 1241.2 1273.2
- FOMCActualVSExpected 2 1238.0 1274.0
- PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1236.2 1274.2
- FridayCloseComparedToThursdayCloseChangeInPercent 1 1236.3 1274.3
- fridayWasATripleWitchDay 1 1237.3 1275.3
<none> 1236.1 1276.1
- FridayIntradayChangeInPercent 1 1238.2 1276.2
- NFPDayOrNot 1 1238.3 1276.3
- didFridayCloseHigherThanThursdayClose 1 1239.0 1277.0
- wasFridayIntradayPositive 1 1240.0 1278.0
- CSIOutcome 3 1244.0 1278.0
Step: AIC=1273.13
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + CSIOutcome + FOMCAction + FOMCActualVSExpected +
PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Df Deviance AIC
- FOMCAction 4 1242.3 1270.3
- FOMCActualVSExpected 2 1238.8 1270.8
- PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1237.3 1271.3
- FridayCloseComparedToThursdayCloseChangeInPercent 1 1237.4 1271.4
- fridayWasATripleWitchDay 1 1238.4 1272.4
- NFPDayOrNot 1 1238.9 1272.9
<none> 1237.1 1273.1
- FridayIntradayChangeInPercent 1 1239.3 1273.3
- didFridayCloseHigherThanThursdayClose 1 1240.1 1274.1
- wasFridayIntradayPositive 1 1240.9 1274.9
- CSIOutcome 3 1245.3 1275.3
Step: AIC=1270.35
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + CSIOutcome + FOMCActualVSExpected + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Df Deviance AIC
- FOMCActualVSExpected 3 1245.8 1267.8
- PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1242.5 1268.5
- FridayCloseComparedToThursdayCloseChangeInPercent 1 1242.7 1268.7
- fridayWasATripleWitchDay 1 1243.5 1269.5
- FridayIntradayChangeInPercent 1 1244.2 1270.2
<none> 1242.3 1270.3
- NFPDayOrNot 1 1244.6 1270.6
- wasFridayIntradayPositive 1 1245.5 1271.5
- didFridayCloseHigherThanThursdayClose 1 1245.7 1271.7
- CSIOutcome 3 1250.4 1272.4
Step: AIC=1267.78
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + CSIOutcome + PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO
Df Deviance AIC
- PrecedingWeekFridayVolumePrecedingWeekThursdayVolumeRATIO 1 1245.9 1265.9
- FridayCloseComparedToThursdayCloseChangeInPercent 1 1246.0 1266.0
- fridayWasATripleWitchDay 1 1246.6 1266.6
- FridayIntradayChangeInPercent 1 1247.6 1267.6
- NFPDayOrNot 1 1247.7 1267.7
<none> 1245.8 1267.8
- didFridayCloseHigherThanThursdayClose 1 1248.8 1268.8
- wasFridayIntradayPositive 1 1249.1 1269.1
- CSIOutcome 3 1253.8 1269.8
Step: AIC=1265.88
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
FridayCloseComparedToThursdayCloseChangeInPercent + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + fridayWasATripleWitchDay +
NFPDayOrNot + CSIOutcome
Df Deviance AIC
- FridayCloseComparedToThursdayCloseChangeInPercent 1 1246.2 1264.2
- fridayWasATripleWitchDay 1 1246.7 1264.7
- FridayIntradayChangeInPercent 1 1247.7 1265.7
<none> 1245.9 1265.9
- NFPDayOrNot 1 1248.0 1266.0
- didFridayCloseHigherThanThursdayClose 1 1248.9 1266.9
- wasFridayIntradayPositive 1 1249.2 1267.2
- CSIOutcome 3 1254.0 1268.0
Step: AIC=1264.15
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
fridayWasATripleWitchDay + NFPDayOrNot + CSIOutcome
Df Deviance AIC
- fridayWasATripleWitchDay 1 1247.0 1263.0
<none> 1246.2 1264.2
- NFPDayOrNot 1 1248.2 1264.2
- didFridayCloseHigherThanThursdayClose 1 1249.1 1265.1
- wasFridayIntradayPositive 1 1249.8 1265.8
- FridayIntradayChangeInPercent 1 1250.2 1266.2
- CSIOutcome 3 1254.3 1266.3
Step: AIC=1263.01
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
NFPDayOrNot + CSIOutcome
Df Deviance AIC
- NFPDayOrNot 1 1248.7 1262.7
<none> 1247.0 1263.0
- didFridayCloseHigherThanThursdayClose 1 1249.5 1263.5
- wasFridayIntradayPositive 1 1250.6 1264.6
- CSIOutcome 3 1255.0 1265.0
- FridayIntradayChangeInPercent 1 1251.0 1265.0
Step: AIC=1262.71
WasMondayOpenHigherThanFridayClose ~ didFridayCloseHigherThanThursdayClose +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
CSIOutcome
Df Deviance AIC
<none> 1248.7 1262.7
- didFridayCloseHigherThanThursdayClose 1 1251.2 1263.2
- CSIOutcome 3 1255.9 1263.9
- wasFridayIntradayPositive 1 1252.6 1264.6
- FridayIntradayChangeInPercent 1 1252.7 1264.7summary(forward_model)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
didFridayCloseHigherThanThursdayClose, family = binomial,
data = df8)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.5668 0.1313 4.317 1.59e-05 ***
CSIOutcomeIsShock -0.9661 0.3891 -2.483 0.0130 *
wasFridayIntradayPositive -0.4103 0.2047 -2.004 0.0451 *
FridayIntradayChangeInPercent 0.2150 0.1084 1.983 0.0474 *
didFridayCloseHigherThanThursdayClose -0.2738 0.1786 -1.533 0.1252
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1249.2 on 914 degrees of freedom
AIC: 1259.2
Number of Fisher Scoring iterations: 4non_random_model <- glm(formula = WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
didFridayCloseHigherThanThursdayClose, family = binomial,
data = df8)
summary(non_random_model)
Call:
glm(formula = WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
didFridayCloseHigherThanThursdayClose, family = binomial,
data = df8)
Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.5668 0.1313 4.317 1.59e-05 ***
CSIOutcomeIsShock -0.9661 0.3891 -2.483 0.0130 *
wasFridayIntradayPositive -0.4103 0.2047 -2.004 0.0451 *
FridayIntradayChangeInPercent 0.2150 0.1084 1.983 0.0474 *
didFridayCloseHigherThanThursdayClose -0.2738 0.1786 -1.533 0.1252
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for binomial family taken to be 1)
Null deviance: 1267.2 on 918 degrees of freedom
Residual deviance: 1249.2 on 914 degrees of freedom
AIC: 1259.2
Number of Fisher Scoring iterations: 4why PrecedingWeekFridayMonth should be random effect
df8$PrecedingWeekFridayMonth <- factor(df8$PrecedingWeekFridayMonth, levels = c("Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"))
table(df8$PrecedingWeekFridayMonth)
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
65 64 87 75 71 87 79 84 78 92 68 69 aggregate(WasMondayOpenHigherThanFridayClose ~ PrecedingWeekFridayMonth, data = df8, mean) PrecedingWeekFridayMonth WasMondayOpenHigherThanFridayClose
1 Jan 0.4923077
2 Feb 0.5312500
3 Mar 0.4827586
4 Apr 0.5866667
5 May 0.4929577
6 Jun 0.5747126
7 Jul 0.5696203
8 Aug 0.5595238
9 Sep 0.4615385
10 Oct 0.5869565
11 Nov 0.5441176
12 Dec 0.6231884library(ggplot2)
ggplot(df8, aes(x = PrecedingWeekFridayMonth, fill = factor(WasMondayOpenHigherThanFridayClose))) +
geom_bar(position = "fill") +
ylab("Proportion")
contingency_table_months <- table(df8$PrecedingWeekFridayMonth, df8$WasMondayOpenHigherThanFridayClose)
print(contingency_table_months)
0 1
Jan 33 32
Feb 30 34
Mar 45 42
Apr 31 44
May 36 35
Jun 37 50
Jul 34 45
Aug 37 47
Sep 42 36
Oct 38 54
Nov 31 37
Dec 26 43# Perform the Chi-square test
chi_sq_test_months <- chisq.test(contingency_table_months)
# Output the result
chi_sq_test_months
Pearson's Chi-squared test
data: contingency_table_months
X-squared = 8.5349, df = 11, p-value = 0.6647chi-squared
# Assign new groupings based on seasons
df8$Month <- with(df8, factor(
ifelse(PrecedingWeekFridayMonth %in% c("Mar", "Jan", "May", "Sept"), "Odd", "Even")))
# Create a new contingency table with the season groupings
contingency_table_Month <- table(df8$Month, df8$WasMondayOpenHigherThanFridayClose)
# Perform the Chi-square test with the new grouping
chi_sq_test_Month <- chisq.test(contingency_table_Month)
# Output the result
chi_sq_test_Month
Pearson's Chi-squared test with Yates' continuity correction
data: contingency_table_Month
X-squared = 3.2023, df = 1, p-value = 0.07353library(lme4)Loading required package: Matrix
Attaching package: 'Matrix'The following objects are masked from 'package:tidyr':
expand, pack, unpackmixed_model <- glmer(WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
wasFridayIntradayPositive + FridayIntradayChangeInPercent +
didFridayCloseHigherThanThursdayClose +
(1 | Month),
family = binomial, data = df8)
summary(mixed_model)Generalized linear mixed model fit by maximum likelihood (Laplace
Approximation) [glmerMod]
Family: binomial ( logit )
Formula:
WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock + wasFridayIntradayPositive +
FridayIntradayChangeInPercent + didFridayCloseHigherThanThursdayClose +
(1 | Month)
Data: df8
AIC BIC logLik deviance df.resid
1260.9 1289.9 -624.5 1248.9 913
Scaled residuals:
Min 1Q Median 3Q Max
-1.6045 -1.0395 0.7570 0.9028 1.6786
Random effects:
Groups Name Variance Std.Dev.
Month (Intercept) 0.008371 0.09149
Number of obs: 919, groups: Month, 2
Fixed effects:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.5359 0.1561 3.433 0.000596 ***
CSIOutcomeIsShock -0.9742 0.3895 -2.501 0.012385 *
wasFridayIntradayPositive -0.4109 0.2050 -2.005 0.044991 *
FridayIntradayChangeInPercent 0.2142 0.1086 1.972 0.048559 *
didFridayCloseHigherThanThursdayClose -0.2733 0.1788 -1.529 0.126307
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Correlation of Fixed Effects:
(Intr) CSIOIS wsFrIP FrICIP
CSIOtcmIsSh -0.051
wsFrdyIntrP -0.449 0.006
FrdyIntrCIP 0.522 0.021 -0.531
ddFrdyCHTTC -0.338 -0.026 -0.399 -0.234library(lme4)
anova(non_random_model, mixed_model)Analysis of Deviance Table
Model: binomial, link: logit
Response: WasMondayOpenHigherThanFridayClose
Terms added sequentially (first to last)
Df Deviance Resid. Df Resid. Dev
NULL 918 1267.2
CSIOutcomeIsShock 1 7.1648 917 1260.0
wasFridayIntradayPositive 1 5.6186 916 1254.4
FridayIntradayChangeInPercent 1 2.8772 915 1251.5
didFridayCloseHigherThanThursdayClose 1 2.3515 914 1249.2#summary(glm(formula = WasMondayOpenHigherThanFridayClose ~ CSIOutcomeIsShock +
# wasFridayIntradayPositive + FridayIntradayChangeInPercent, family = binomial,
# data = df8))#create df8$non_random_model_predict and df8$mixed_model_predict by going through each row of df8
# Predict with the non-random model
df8$non_random_model_predict <- as.integer(predict(non_random_model, type = "response") > 0.5)
# Predict with the mixed model
# Note: For mixed models, if you want to include random effects in predictions, use re.form = NULL
df8$mixed_model_predict <- as.integer(predict(mixed_model, re.form = NULL, type = "response") > 0.5)non_random_model_right_wrong <- c(0, 0)
mixed_model_right_wrong <- c(0, 0)
# Iterate through each row of df8
for (i in 1:nrow(df8)) {
# For the non-random model
if (df8$WasMondayOpenHigherThanFridayClose[i] == df8$non_random_model_predict[i]) {
non_random_model_right_wrong[1] <- non_random_model_right_wrong[1] + 1
} else {
non_random_model_right_wrong[2] <- non_random_model_right_wrong[2] + 1
}
# For the mixed model
if (df8$WasMondayOpenHigherThanFridayClose[i] == df8$mixed_model_predict[i]) {
mixed_model_right_wrong[1] <- mixed_model_right_wrong[1] + 1
} else {
mixed_model_right_wrong[2] <- mixed_model_right_wrong[2] + 1
}
}
# Print the results
print(non_random_model_right_wrong)[1] 509 410print(non_random_model_right_wrong[1] / (non_random_model_right_wrong[1] + non_random_model_right_wrong[2]))[1] 0.5538629print(mixed_model_right_wrong)[1] 519 400print(mixed_model_right_wrong[1] / (mixed_model_right_wrong[1] + mixed_model_right_wrong[2]))[1] 0.5647443# save df to .csv
# Save df8 to a CSV file
write.csv(df8, file = "results/SPY_non_mixed_and_mixed_model.csv", row.names = FALSE)